부산대학교 도서관

This study investigates the feasibility of using the cross-correlation between mean arterial blood pressure (MABP) and mean cerebral blood flow velocity (MCBFV) as a signature to distinguish diabetics with various degrees of autonomic neuropathy. 54 subjects were recruited. Among them, 15 were healthy adults (normal subjects), 17 were diabetics with mild autonomic neuropathy symptoms, and 22 were diabetics with severe autonomic neuropathy symptoms. The cross-correlation function of pre-filtered spontaneous MABP and MCBFV was computed. The maximum peak and the corresponding standard deviation along with the maximum peak index and the corresponding standard deviation were extracted from three frequency ranges, namely very low frequency (0.015-0.07Hz), low frequency (0.07-0.15Hz), and high frequency (0.15-0.40Hz), in the supine position, giving a total of 12 features for each subject. After feature reduction using a greedy forward selection technique, each subject was classified into one of three possible classes. The feature vectors were classified based on a support vector machine (SVM) classifier with a classification rate of 90.74%. The results indicate that SVM classification based on the cross-correlation between MABP and MCBFV could be an effective approach for discriminating diabetics with various degrees of autonomic neuropathy.